Given an infinite series of Bernoulli RVs $X_1,X_2,...$ (which may be differently distributed and mutually dependent), we are given that for every $n>0$, $$\mathbb{E}\left[\sum_{t=1}^{n}(1-X_t)\right] \leq a\log n~,$$ with $a>0$ some deterministic parameter. Intuitively, this means that when $n$ is large enough, the expected number of failures until time $n$ becomes negligible.
My question is whether this implies a bound on $\mathbb{E}\tau$ as a function of $a$ (I will be glad as long as $\mathbb{E}\tau$ is guaranteed to not be infinite), where $$\tau \triangleq \min \{t>0:X_t=1\} $$ is the time of the first success.
Thank you!
Edit: Sorry for the confusion in my original post; I changed it to clarify that the bound on $\mathbb{E}\tau$ should not depend on $n$ and that $\tau$ stands for the first success at any time.